AI Shouldn’t Be Trusted: Why Understanding That Can Be Transformative

Amidst the hype surrounding AI, there's a growing realization that AI isn't as "smart" as we've been made to believe. TechCrunch's Devin Coldewey describes AI as "the world's most comprehensively informed bullshitter rather than an authority on any subject."

As a CTO, one of key elements of success is to be able to understand technology. The way I see it is that at a given point in time, you have a technology and a set of problems. And for that given technology there is a boundary: inside that boundary are the set of problems it solves well, and outside that boundary are the set of problems it doesn’t work for.

Right now with AI, we’re seeing just how important it is for CISO’s to understand technology.

CTOs: Be intentional about finding a tech’s true value

A key function that CTOs execute is looking at a technology and clearly see the shape and size of that boundary . This is not something you can do on your own; you have to read and talk to people who help you discover and see that boundary clearly. But this also comes with challenges.  

On one hand, you have the folks who are educators. Like you, they try their best to say what’s good and bad about a technology and try to shine a light on parts of that boundary. On the other hand, you have those that are trying to influence your decision, and they just want to tell you that tech solves many more problems than it does – or sell the promise of that. Even if you assume positive intent, the lack of clarity can be damaging and lead you as a technology leader to make incorrect decisions that could have negative business impacts.  

CTOs need to seek and find truth, and only out of that truth can they harness the technology for the good of the business. Without that, they are doomed, their results just being a function of luck, not skill.

Why the AI tech boundary is smaller than it appears

So now to artificial intelligence (AI), or this recent generative AI wave. I think what is being done is truly amazing and transformative. But I see we are being influenced to think that the AI “tech boundary” is much larger than it is today. That incorrect expansion is clouding our judgment and could cause you problems identifying good solutions to buy or build.

We’ve all seen AI hallucinating, and we’ve seen the Mijourney pictures of the people with 6 fingers and 50 teeth. It’s kind of creepy. I mean 5 or 6 fingers, that’s just an unimportant detail right? I even happened to ask ChatGPT who was the CTO of Illumio – and I have been for 10 years, so that should be in its body of knowledge – and according to it I’m not…hmmm.

Trimming AI’s tech boundary reveals its real value

Here’s where I start to find it exciting. Take perfection and put it squarely outside the AI tech boundary. Take high accuracy and precision and put them outside the AI tech boundary. Now consider what’s inside and what innovations can focus on problems where you only have to be 50 percent right.  

Devin Coldewey had a great quote in his TechCrunch article that illustrates the same point: “All that matters is that these systems do not distinguish between something that is correct and something that looks correct. Once you understand that the AI considers these things more or less interchangeable, everything makes a lot more sense.”  

What are some of those things? I do think generative art is a place that has a lot of potential for AI. Art is subjective. What speaks to one person doesn’t to another. And there are lots of times that “good enough” is sufficient for your need, as there could be hundreds of really good images that could be shown on your presentation to communicate your point.  

Another great example is the problem we all seem to have staring at blank sheet of paper. When I sit down to write a blog entry (including this one), it’s a hurdle to get going. Generative AI could write a first draft, and we could call it the zeroth draft. Even if you took nothing from what the AI created or if it served as a catalyst for getting started, it still would be valuable.

Those are just two examples of cases where just being 50 percent right provides value. I’m sure there are many more, and I’m in search of these kinds of cases that are inside the true AI “tech boundary” in my own technology exploration.  

Know tech’s true boundaries

Consider making sure you understand the true “tech boundary” as you chase your creative and innovative endeavors – there is a sweet spot to find today.  

I would love to hear from those who agree (or disagree) on ideas of the AI tech boundary. Share your thoughts with @illumio on LinkedIn or Twitter.